Passive optical networks (PONs) have been deployed worldwide. A network roll-out requires significant investment, and once constructed, should be able to be upgraded economically. Recent activities in both the Full Service Access Network (FSAN) organization, and the IEEE 802.3av 10G-EPON (Ethernet Passive Optical Network) study group have studied next-generation access networks. Both organizations have considered PONs (Passive Optical Networks) operating at 10Gb/s (10 Gigabits-per-second—“10G communication”) rates. It would be desirable to have an upgrade which can leverage the existing infrastructure without requiring outside plant adjustment or changes in customer premises equipment.
The 10G-EPON group has specifically addressed the issue of “coexistence” of a new 10G-EPON standard with an existing GE-PON (Gigabit Ethernet Passive Optical Network). By using a combination of Wavelength Division Multiplexing (WDM) and Time-Division Multiple Access (TDMA), it is possible for both a legacy 1G PON and new 10G PON to operate on the same network.
This scheme is illustrated schematically in FIG. 1. Network 100 may include 1G/10G Optical Line Terminal (OLT), dual rate polarization mode dispersion (PMD) device 110, 1G Optical Network Unit (ONU) 104, second 1G ONU 106, and 10G ONU 108. OLT 102 includes a Media Access Control (MAC) which controls both the legacy 1G ONUs 104, 106 and new 10G ONU 108.
A large link budget in the downstream direction is not a problem since the OLT 102 can use either a high power Distributed Feedback Laser (DFB) laser or a DFB laser in combination with an SOA (Semiconductor Optical Amplifier) to meet the link budget. In both the specifications for (Ethernet Passive Optical Networks—EPON—IEEE 802.3ah) and for (Gigabit capable Passive Optical networks—GPON—ITU-T G.984), the upstream wavelength is defined to range between 1260 and 1360 nm (1310+/−50 nm) and the downstream wavelength is defined to lie within the 1480-1500 nm band. (1490+/−10 nm). With regard to 10G communication, the 1571 nm wavelength was the working wavelength chosen by the IEEE 10G-EPON study group (IEEE 802.3av) for 10G downstream communication, that is communication from OLT 102 to various 10G ONUs, such as ONU 108.
Thus, one limiting performance factor in the system of FIG. 1 occurs in the upstream communication direction (that is, toward the OLT 102). The optical powers of the legacy 1G ONU 104, 106 transmitters are fixed, while it is desirable to use a lower power transmitter in the 10G ONU 108 to keep system costs to a minimum. The challenge for the OLT 102 receiver is meeting a large (e.g. 29 dB) link budget for both the 1G and 10G data communication streams. Legacy 1G ONUs 104, 106 must be used, and the 1G signals operate over a specified wavelength range of 1260-1360 nm (nanometers), without Forward Error Correcting (FEC), in the upstream direction (toward the OLT 102). Since the 10G ONU PMD is not yet defined, the wavelength may be specified with more precision within the 1260-1360 window, and FEC may be available to meet required link budgets. Some existing approaches are discussed below.
There are two basic approaches to using an Avalanche Photo-Diode (APD) as a dual-rate receiver, which are illustrated in FIGS. 2 and 3, respectively. One goal when using an APD at both 1G and 10G rates is optimizing the bandwidth of the Trans-Impedance Amplifier (TIA).
FIG. 2 shows network 200 which includes APD 202, resistance 204, TIA 206, 1G Band Pass Filter (BPF) 208, and 10G BPF 210. The resistance value “R” of resistance 204 is the resistance across the TIA 206. The bandwidth of the TIA 206 will be proportional to 1/R while the (thermal noise current)2 of the receiver will also be proportional to 1/R. If the bandwidth of the TIA 206 is large enough to enable throughput of both 10 Gb/s (10G) and 1.25 Gb/s (1G) signals, then the thermal noise current of TIA 206 will be (10/1.25)1/2, or 4.5 dB higher than under ideal circumstances for the 1.25 Gb/s signal.
For thermal-limited receivers (such as PIN, and APD to a lesser extent), the receiver sensitivity is proportional to 1/(thermal noise current). With such a static TIA, the 1G sensitivity in the dual-rate receiver will be 4.5 dB lower (i.e. worse) than in an APD-TIA combination optimized for 1G operation.
One potential solution to minimize this penalty is to vary the value of “R” in time such that the TIA 206 bandwidth is optimized for either 1G or 10G traffic. However, this approach adds significant complexity, as the receiver must implement a high-speed dynamic TIA 206 as well as communicate with the MAC in OLT 102 to track the bit-rate of the incoming signals.
Another approach is to use two separate APDs, as shown in FIG. 3. FIG. 3 is a block diagram of network 300 that includes a 3 dB optical splitter 302 (labeled with “3 dB” in FIG. 3), APDs 304, 306, a 1G TIA and BPF 308, and a 10G TIA and BPF 310. However, with this approach, performance suffers due to the insertion loss of the optical splitter 302, which effectively halves the power that is directed along each of the two branches to the right of splitter 302.
In the following, it is noted that sensitivities having measurements in “dBm” having larger negative numbers are most beneficial for operation of an optical communication network.
Good APDs provide −34 dBm sensitivity with a Bit Error Rate (BER) of 10−12 at 1G and −25 dBm sensitivity at 10G. A single APD having the foregoing specifications, if used as a dual-rate receiver, in the network of FIG. 2 would yield a sensitivity of −29.5 dBm at 1G and −25 dBm at 10G, while providing −31 dBm and −22 dBm sensitivities for the 1G and 10G data streams, respectively, if deployed within the network shown in FIG. 3.
The sensitivity disparities between the 1G and 10G data communication streams described above are undesirable. The performance of the networks of FIGS. 2 and 3 could result in the 1G communication operating well, and the 10G communication malfunctioning to an unacceptable degree. Accordingly, there is a need in the art for an improved system and method for receiving data at multiple data rates.